AI Debt Boom: Lessons From the Telecom Bust

AI debt boom compared with the telecom bust, shown with data center and fiber imagery

Growth-debt lender David Spreng, co-CEO of Runway Growth Capital, sees the AI debt boom as a repeat of the telecom bust. His argument is that AI financing is moving from equity to bonds and private credit, and that debt punishes a wrong demand forecast far harder than equity does.

From equity to debt

Until recently the AI buildout was mostly paid for with venture equity. Global venture funding hit $679 billion in the first nine months of 2026, the most ever for that stretch. AI companies took $102 billion in the third quarter alone, roughly 64% of all venture dollars. Twenty-seven companies, a record, closed rounds of $1 billion or more, together about a third of the total.

Equity has no fixed repayment date. When an investment underperforms, valuations fall and the company keeps room to pursue its plans. Debt doesn’t give that room.

Debt is now showing up in volume:

  • Meta sold $30 billion of bonds last October and returned in April for $25 billion more, with investors reportedly demanding better terms the second time.
  • Amazon, Alphabet and Oracle have each raised tens of billions in debt.
  • A joint-venture partner raised a record $27 billion bond to pay for Meta’s Louisiana data-center campus.
  • Morgan Stanley projects roughly $3 trillion of infrastructure spending by the big AI and cloud companies through 2028, with only half coming from their own cash flow.

The rest, Spreng says, has to come from the bond market and private credit.

What the telecom bust looked like

Spreng lived through the earlier cycle, first as a late-1990s venture capitalist and later lending to growth-stage tech firms. Telecom companies borrowed heavily on a forecast everyone repeated and few checked: internet traffic doubling every 100 days.

In reality traffic doubled about once a year. Strong growth, but a fraction of what lenders had priced in. By 2002 the sector owed roughly $1 trillion, much of which the FCC chairman told Congress would never be repaid, and WorldCom filed what was then the biggest bankruptcy in US history.

Much of the fiber laid then sat unused for years. Buyers who got it for pennies on the dollar in bankruptcy finally lit it up, after a decade of restructurings and mergers that rebuilt entire cap tables. Some companies survived. Many didn’t.

The demand story today

The echo that worries Spreng most is the demand narrative. The current version says data centers are effectively leased before they’re built, and that building too little is the only real mistake. That may prove right. But he notes it’s repeated with the same uniformity as the 1999 traffic forecast, and much of it rests on projections from the companies doing the building.

Anthropic’s public listing may start to test it. Once companies of that scale report results every quarter, investors can compare the demand story with audited numbers.

The counterargument

Spreng grants the opposing case. The biggest borrowers may be among the most profitable companies ever built, with cash flows telecom carriers never had. A few AI companies also seem to be scaling revenue faster than any business before. And unlike the fiber boom’s traffic forecast, the internet’s growth did arrive.

His objection is timing. A lender must ask whether cash comes in when the debt falls due. Equity can sit through a wrong timeline and debt can’t, so if borrowed money builds capacity faster than revenue can pay for it, he expects restructurings instead of quiet markdowns.

What lenders will be asking

Spreng isn’t predicting that AI will end like telecom. He says the lasting lesson of the bust was about capital structure, which decided who was still standing when demand caught up with capacity years later. He has long said growth companies are too wary of debt, and he calls that view conditional.

In his framing, debt works when payments are sized to revenue a borrower can already see and lenders underwrite fundamentals instead of narratives. It treats everyone else harshly. Quarterly reports from newly public AI companies and the pricing of the next big bond sales will show which side of that line the borrowing sits on.

Source: Crunchbase News

Written by

Saksham Bhargava

Saksham Bhargava runs Alteroid and has been writing about consumer technology for over five years, covering iPhone, Android, Windows, AI tools and gaming. With an MCA and six years of hands-on web development experience, he tests the features he writes about on his own devices and focuses on clear, practical guides that solve real problems. Away from the screen, he enjoys travelling to places most tourists skip.